Chemical transfer during redox exchanges between H-2 and Fe-bearing silicate melts.

HAL Id: hal-00069454

https://hal-insu.archives-ouvertes.fr/hal-00069454

Submitted on 31 May 2006

HAL

is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire

destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Chemical transfer during redox exchanges between H-2 and Fe-bearing silicate melts.

Fabrice Gaillard, Michel Pichavant, Stephen Macwell, Rémi Champallier, Bruno Scaillet, Catherine Mccammon

To cite this version:

Fabrice Gaillard, Michel Pichavant, Stephen Macwell, Rémi Champallier, Bruno Scaillet, et al.. Chem-

ical transfer during redox exchanges between H-2 and Fe-bearing silicate melts.. American Mineralo-

gist, Mineralogical Society of America, 2003, 88, pp.(2-3) 308-315. �hal-00069454�

0003-004X/03/0203–308$05.00 308

INTRODUCTION

In concentrations ranging from parts per million to weight percent, hydrogen is always present in natural minerals or melts in the form of a variety of species with different redox states (H⁺, OH^–, H2O, or H2) (Ingrin and Skogby 2000; Johnson et al.

1994; Schmidt et al. 1998; Stolper 1982). The identification of the nature and the mobility of the different H species is a key for understanding the Earth because H incorporation in min- eral or melt dramatically modifies their physical properties. In nominally anhydrous minerals (NAM), the kinetics and mecha- nisms of H incorporation or extraction have been and still are studied intensively. Water-derived species incorporation/extrac- tion in NAM recently has been revealed to evolve through a two-stage process (Kohlstedt and Mackwell 1998). The fastest stage was demonstrated to be related to redox processes in- volving H motion as protons and electronic defects related to Fe³⁺-Fe²⁺ exchange (Kohlstedt and Mackwell 1998; Hercule and Ingrin 1999). Therefore, both the H content and Fe³⁺/Fe²⁺ ratio in NAM are strongly interdependent parameters. The slower stage does not involve redox exchanges and seems to be asso- ciated with intrinsic defect mobility (Kohlstedt and Mackwell 1998).

With respect to silicate melts, it is generally accepted that

H-bearing species diffuse as H2O molecules (e.g., Zhang et al.

1991; Behrens and Nowak 1997). However, at low water con- tents (<0.8 wt%) the mobile water-derived species may be NaH2O⁺, H3O⁺, or H⁺ (Stanton et al. 1990). Schmidt et al. (1998) have also identified the possibility of hydrogen incorporation as H2 moleculesin Fe-free silicate melts. In natural melts, which all contain H and Fe, redox interactions occur between H2-H2O and Fe³⁺-Fe²⁺ (Gaillard et al. 2002). Currently, the mechanisms and the mobilities of the species involved in these interactions are still not known accurately.

Gaillard et al. (2002) studied the kinetics of Fe redox reac- tions at 2 kb in H2O-rich (5–6 wt%), Fe-poor (1–3 wt% FeOtot) melts that occur in response to variations of hydrogen fugacity (f_H2). No redox fronts were observed. The kinetic data were interpreted in terms of a two-step reaction mechanism that in- volves first a virtually instantaneous diffusion of H2 in the sample and then slower structural/chemical reorganizations, involving slower interactions of water-derived species with Fe in the melt. Oxidation-reduction of Fe in these low-Fe, high- H2O rhyolitic melts is reaction-limited, in contrast to the diffu- sion-limited process identified by Cooper et al. (1996) for the oxidation of dry basaltic melts.

From these two studies, it is clear that the redox mecha- nisms in silicate melts strongly depend on the presence or ab- sence of hydrogen (as H2 and OH/H2O). In addition, another major difference concerns the Fe concentration of the studied melts. Both factors make direct comparisons between the two studies difficult.

* Present address: Bayerisches Geoinstitut, Universität Bayreuth, D-95440 Bayreuth, Germany. E-mail: fabrice.galliard@uni- bayreuth.de

Chemical transfer during redox exchanges between H

and Fe-bearing silicate melts

BRUNO SCAILLET,¹AND CATHERINE MCCAMMON²

1CNRS, ISTO- 1 A rue de la Férollerie, Orléans cedex 02 45071, France

2Bayerisches Geoinstitut, Universität Bayreuth, D-95440 Bayreuth, Germany

ABSTRACT

Kinetics and reaction paths of Fe³⁺ reduction by H2 in high-Fe and low-H2O silicate melts have been investigated at 800 ∞C. Time-series experiments were performed in cold-seal pressure vessels at 50 bars of pure H2 using rapid-heating and rapid-quench strategies. Within the first minutes of the experiments, a fast partitioning of Na occurred between the gas and the melt due to the reducing conditions. Kinetically decoupled from the Na partitioning, the progression of a front of Fe³⁺ reduc- tion within the quenched melt was observed and was identified as a diffusion-limited process. The growth of the reduced layer is accompanied by an increase in concentration of OH-groups suggest- ing that reduction operates through proton incorporation within the melt. As this growth rate is slightly faster than predicted from the diffusion of molecular H2O, a different and mobile water- derived species seems likely. One possible mechanism is the reduction of Fe³⁺ by the transport of molecular H2. As this process is limited by the flux of H2, it will depend on both diffusivity and solubility of H2 in the melt. Alternatively, migration of protons (H⁺) and electronic species within the melt could control the velocity of the reduction front. The increase in concentration of the reaction- derived OH groups produces a water over saturation followed by partial dehydration of the melt.

This dehydration leads to a change in the redox conditions within the gas that influences the Na partitioning between gas and melt.

GAILLARD ET AL.: REDOX EXCHANGES BETWEEN H₂ AND IRON IN MELTS 309

In this work, reduction experiments have been performed in a near-pure H2 atmosphere on a nearly anhydrous Fe-rich silicate melt at low pressures to help elucidate the role of H2 on redox mechanisms in the low-water concentration range. We show that, in contrast to previous observations for Fe-poor and high-H2O melts, a redox front is clearly identified. We charac- terize the rate law for the progression of this redox front and the associated chemical mass transfers in the melt. We demon- strate that the process that rate-limits reduction is much slower than rates of molecular H2 diffusion in the Fe-free system, but is slightly faster than molecular H2O diffusion. Therefore, this study brings evidence that different water-derived species with different apparent mobilities may exist in silicate melts.

EXPERIMENTALTECHNIQUES

Starting glass

Crystal-free, natural peralkaline obsidian was used as a start- ing material (Eburru, Kenya; MacDonald and Bailey 1973).

The composition of the glass, including Fe and Fe2O3 (25% of iron as Fe³⁺) and water content (0.25 wt% H2O), is provided in Table 1. A series of 11 water content determinations, performed by FTIR (see Analyses), on different wafers yielded 0.25 ± 0.01 wt% H2O, suggesting a homogeneous distribution of water-derived species in the starting glass.

The strong peralkalinity of this obsidian places the glass transition temperature at 400–500 ∞C, and allows us to per- form low-temperature experiments under nearly dry conditions in the melt stability field without crystallization. A dry glass of the albite-orthoclase-quartz (AOQ) system was also used in one experiment as an Fe-free reference (Table 1).

Materials and strategy

Cylinders of the Eburru glass were placed in tubes of pure gold (ID = 5 mm, OD = 5.4 mm; the glass cylinder was cored using a drill of 5 mm ID). These assemblies, with one end of the capsule arc-welded and the other end open, were placed in cold-seal pressure vessels at 800 ∞C under 50 bars of pure H2

for various durations (see Table 2). One additional experiment was performed using a platinum-capsule with one end sealed by an AOQ glass cap and the other end open to H2 (Fig. 1a).

The AOQ glass cap was formed by melting glass powder in the platinum-capsule at 1400 ∞C and 1 bar for 6 hours. A cylinder of Eburru glass was then placed in the same capsule in contact with the AOQ glass. This assembly was heated to 1200 ∞C at 1 bar for 7 minutes to make an airtight contact between the two glasses and between the samples and the capsule wall. The end of the capsule with the AOQ glass was cut so that the AOQ glass was directly exposed to the H2 gas in the experiment, and thus behaved as a screen between one end of the Eburru glass and the H2 atmosphere. The other end of the Eburru glass was

in direct contact with H2.

The vessels used in the experiments allow rapid heating and quenching to be performed by mechanically moving the sample rapidly while at pressure between the cool and hot parts of the vessel. Once the experimental temperature was attained (~30 min), the samples were moved to the hot part of the vessel.

Using this strategy, each sample was heated from ambient tem- peratures to 800 ∞C within less than 5 min. After the experi- ment, the sample was quenched rapidly by moving it to the cool part of the vessel. Subsequently, the samples assemblies were cut into slices and polished for optical observation, elec- tron microprobe analysis (EMPA), and infrared spectroscopy (FTIR). Wet-chemical analyses and Mössbauer spectroscopy were also performed on selected samples.

Analyses

EMPA was performed using an SX-50 Cameca electron microprobe under the following conditions: 15 kV accelerat- ing potential, 6 nA beam current, 10–20mm beam size, 10 s counting time on element peak positions and 5 s counting time on the background. Multi-element chemical profiles were per- formed with a 20 mm step increment across the sample.

FTIR analysis was performed on doubly polished samples using a Nicolet 760 Magna spectrometer equipped with an in- frared microscope. A CaF2 beamsplitter was used with a vis- ible light source and liquid N2 cooled MCT/A detector. The mean beam size was 80–100 mm. Total H2O content was esti- mated from the height of the 3600 cm^–1 stretching O-H band.

The extinction coefficient of the 3600 cm^–1 absorption band was calibrated against measurements using Karl-Fisher titra- tion performed on the starting glass (Table 1). The extinction coefficient was determined to be 40 L/g/cm.

Wet-chemical techniques were used to determine the FeO content of selected samples. The method and its precision are detailed in Gaillard et al. (2001). Fe2O3 contents were obtained by subtraction of FeO from total Fe analyzed by EMPA. As wet chemistry is a bulk technique, only the samples interpreted as chemically homogeneous based on their uniform color were analyzed (samples Eb/0, Eb/6). In addition, analyses using the Mössbauer milliprobe (McCammon et al. 1991; McCammon 1994) were performed on sample no. Eb/5 with a spot size of

~300 mm to determine Fe³⁺/Fe^tot in the two regions of the sample with different colors.

RESULTS

Optical observation

The glasses quenched from all time-series experiments showed a sharp contrast in color, having translucent-green rims and inner regions with the same dark color as the starting glasses (see Fig. 2). The boundary between translucent-green and dark

TABLE 1. EMP analyses of the starting Eburru Obsidian and the AOQ glass

Sample SiO2 TiO2 Al2O3 MgO FeOtot MnO NiO CaO Na2O K2O SO3 P2O5 Total

Eburru 70.13 0.28 8.16 0.02 6.84 0.23 0.04 0.21 6.48 4.40 0.01 0.02 96.81

Eburru FeO = 5.31wt%, Fe2O3 = 1.7 wt%, H2O = 0.25 wt%

AOQ 70.10 0 12.10 0 0.02 0 0 0.03 4.40 5.10 0 0 98.75

Notes: Wet chemistry was used for FeO and Fe2O3 contents and H2O was determined by Karl Fisher titration. All data are given in wt%.

zones moved toward the inner part of the sample as the experi- ment duration increased. The rate of progression of this bound- ary was characterized optically. Table 2 summarizes the 6 observations from the 6 different run durations. After a run of 2916 min (~49 h), the sample was completely translucent-green (Eb/6). Micro-crystals were observed within the glasses after the runs. Using optical techniques we identified feldspar crys- tals, but these were not analyzed with the EMP.

Fe³⁺/Fe^tot

Wet chemistry results, expressed in terms of Fe³⁺/Fe^tot, are presented in Table 2. The starting obsidian (Eb/0) has Fe³⁺/Fe^tot

= 0.23 ± 0.05, whereas sample no. Eb/6 was characterized as an essentially Fe³⁺-free glass (Fe³⁺/Fe^tot = 0.01 ± 0.01). The two micro-Mössbauer analyses were performed on sample no. Eb/

5 at the positions shown in Figure 2. The dark zone yielded Fe³⁺/Fe^tot = 0.25 ± 0.05, similar to the starting obsidian (Eb/0), Reduced glass: Pt-capsule: AOQ glass:

Oxidized glass: 1 mm

Eburru

H2

EBURRU

AOQ

A H2

H2

AOQ

H2

EBURRU

gl d

-0.05 -0.03 -0.01 0.01 0.03 0.05 0.07 0.09 0.11 0.13 0.15

2300 2600 2900 3200 3500 3800 4100 4400 4700 5000

Waves numbers (c m^-1)

Absorbance units

H2

H₂O AOQ- Run Eb/1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

2300 2600 2900 3200 3500 3800 4100 4400 4700 5000

Waves nu mb er cm-1

Absorbance units

Eb/1- Eburru ass Redu ced an ox yd ized zones

H2

B

FIGURE 1. Sketch of the albite-orthoclase-quartz + Eburru glasses + Pt-capsule assembly used for experiments no. Eb/1. (A) Before experiments, (B) After experiments. FTIR spectra of AOQ (note the observed H2-band) and both reduced and non-reacted zone of the Eburru glass are also shown. See text for additional details.

GAILLARD ET AL.: REDOX EXCHANGES BETWEEN H₂ AND IRON IN MELTS 311

whereas no Fe³⁺ was measurable in the green zone (Table 2).

This observation gives credence to the argument that the dif- ference of color between the green and dark zone indicates a strong difference of Fe redox state. Given the strongly re- ducing conditions of the experiment and the homogeneity of color within the green zone, we propose that the Fe³⁺ con- tent is constant in this region and nearly equal to 0 for all runs (Table 2).

Major elements

EMP chemical profiles were performed on each experimen- tal charge. In the run products no. Eb/1, no. Eb/2, and no. Eb/3, no clear major-element concentration variations were observed across the glasses. Comparisons with the starting glass revealed that Na content decreased by about 7% for all of these run prod- ucts (Table 2). For samples no. Eb/4 and no. Eb/5, gradients in the concentration of Na at the samples rims were observed (Fig.

3). The maximum in Na concentration located at the gas-melt

interface was equal to the Na content of the starting glass. Sur- prisingly, the position of this chemical gradient does not match that of the redox boundary. Rather, it is located behind the re- duction front. Thus, for run no. Eb/5 (Figs. 2 and 3), the re- duction front is at 480 mm from the sample surface, whereas the first evidence of Na motion was identified at 150–180 mm. No motion of other major elements was detected by EMPA (Fig. 3). For sample no. Eb/6, the Na2O distribution within the sample was homogeneous and matched that of the starting glass.

H2O

FTIR profiles were performed for several samples (see Table 2, and Figs. 2 and 3). In all samples analyzed, we observed water to be present mainly in the form of OH groups due to the dominance of the 4500 cm^–1 O-H band over the 5200 cm^–1 H2O band. The green zone has significantly higher water contents than the dark zone (0.73–0.43 vs. 0.25 wt% H2O, respectively).

In the dark zone, the water content was similar to that of the starting glass. Between the green and the dark zones, the water content seems to decrease sharply although the spatial resolu- tion of the FTIR analysis does not allow precise determination of the shape of the water profile at this boundary. The H2O content of the green zone is not homogeneous. Rather, a maxi- mum H2O concentration is reached near the green/dark bound- ary (0.73 wt%) and it drops to ~0.43 wt% at the sample rims.

For the longest experiment, no. Eb/6, the color and the chemi- cal composition including H2O content are homogeneous. The uniform H2O content of 0.42 wt% is similar to the values mea- sured at the edges of the other samples (Table 2).

H2 mobility in melts

Experiment no. Eb/1, where the sample adjoined the AOQ glass, was characterized by growth of reduced green layers that are similar in width on the sides directly exposed to H2 and in contact with the AOQ glass (Fig. 1). FTIR spectra were col- TABLE 2. Summary of the run conditions and descriptions of starting and run products

Experimental: 800∞C, f_H2 = 50 bar

Sample name Eb/0* Eb/1 Eb/2 Eb/3 Eb/4 Eb/5 Eb/6

Run duration

(Minutes) 0 30 65 90 150 180 2916

Capsule – Pt+AOQ Au Au Au Au Au

Optical

Advancement of the 0 180 302 355 430 480 >1800

reduction front in mm

Fe³⁺/Fetot (Wet chemical analyzes no. Eb/0 and Eb/6; Mössbauer analyzes no. Eb/5, see figure 2)

Green zone none n.a n.a n.a n.a 0 0

Dark zone 0.23 n.a n.a n.a n.a 0.25 none

FTIR (H2O wt%) Green zone

External† none 0.66 n.a 0.57 0.41 0.43 0.42‡

Middle† none n.a n.a 0.69 0.67 0.42‡

Internal† none 0.75 0.74 0.73 0.71 0.73 0.42‡

Dark zone (homogeneous) 0.25 0.23 0.24 n.a 0.26 0.25 none

Notes: n.a. = Non analyzed; None = Not visible within the sample.

* Starting glass.

† Position relative to the sample geometry (see figure 2 for the location).

‡ 6 FTIR points were measured on this sample indicating a homogeneous H2O content (±0.04 wt%).

FIGURE 2. Transmitted light photomicrograph showing the change in color of the glass associated with reduction of Fe³⁺. The two large circles indicate the location of the Mössbauer milliprobe analyses listed in Table 2. The five small circles illustrate the location of FTIR analyses from Table 2. Charge no. Eb/5.

lected on the AOQ sample before and after the experiment.

The water content was about 0.01 wt% before and 0.015 wt%

H2O after the experiment (see Fig. 1). A striking peak at 4115 cm^–1 was observed everywhere in the AOQ glass. This peak, which is always observed after annealing under very high f_H2 conditions (Schmidt et al. 1998), is due to molecular H2 dis- solved in the glass. This observation establishes unambiguously that, within a very short time, molecular H2 had permeated the entire AOQ glass whose thickness is about 4 mm. No H2 peak was observed for the Eburru glass in this experiment.

DISCUSSION

In these experiments, several reactions operate as suggested by the decoupling between H2 mobility, Na migration and the reduction rate of Fe³⁺. For clarity, hereafter we discuss sepa- rately the different processes.

Constraints on the Fe³⁺ Æ Fe²⁺ rate constant

Figure 4 shows a plot of the square of the reduction front position as a function of time. The linear relationship suggests that the reduction of Fe³⁺ to Fe²⁺ is a diffusion-limited process.

The calculated slope gives the parabolic rate constant for Fe³⁺

Æ Fe²⁺ at 800 ∞C, KFe3+ Æ Fe2+ = 2 ¥ 10^–11 m²/s. Chekhmir et al.

(1985) also observed the progression of a reduction front in albite melt (claimed to be Mn⁷⁺ Æ Mn²⁺), which they postu- lated to be controlled by diffusion and internal decomposition of molecular H2. We calculated a parabolic rate constant from their single experiment and we find KMn7+Æ Mn2+ = 2 ¥ 10^–10 m²/s at 1000 ∞C, which seems reasonably consistent with our lower temperature results. However, because Chekhmir et al.

(1985) did not perform systematic time-series experiments, the comparison is tenuous. It is interesting to note that, in both studies, the parabolic rate constant is close to but slightly higher than the diffusion of molecular water [(~ 0.5 to 0.8 log unit higher, D_H2O calculated after Zhang et al. (1991)]. For run no.

Eb/5, the coupled incorporation of H plus the reduction of Fe³⁺

have progressed over 480 mm (Table 2), whereas in the same time, migration of molecular H2O should have operated over

~220 mm. Thus, we have obtained evidence that redox interac- tion between H2 and Fe gives rise to a transfer of water-derived species with mobilities different from molecular H2O diffusion.

Constraints on H2 mobility from no. Eb/1

The presence of molecular H2 in the AOQ glass, and the similarity between the widths of the reduction rim at the H2- melt and melt-AOQ interfaces, suggest that H2 has diffused very rapidly through the AOQ melt (run no. Eb/1). We excluded the possibilities of H2 migration at the AOQ-Eburru boundary because we consider the interface to be welded tightly in the previous anneal. We estimate that if H2 had taken more than 5 min to cross the 4 mm length of the AOQ cap, we should have been able to see asymmetry in the width of the reduction rim.

Therefore, we can calculate a minimum diffusion coefficient for molecular H2 in the AOQ melt using the relationship x = (D·t)^1/2, where x is the diffusion length (4 mm), D the diffusion coefficient, and t the minimum duration required for diffusion across the cap. We estimate D_H2≥ ^{10–8 m2}/s at 800 ∞C. This value is significantly higher than the diffusion coefficient pro- posed by Chekhmir et al. (1985) for diffusion of molecular H2

in amorphous albite under essentially the same conditions.

Mechanisms of reduction: Ionic vs. molecular migrations Macroscopically, reduction of a melt results from an increase of the cation/oxygen ratio. This can be accomplished by a loss of oxygen anions or by incorporation of cations (see Fig. 5 for an illustration of all the likely scenarios). The growth of re- duced rim is accompanied by an abrupt increase of the OH content, which is consistent with an incorporation of protons (increase of cation/oxygen ratio). Mechanism 1, shown in Fig- ure 5, involves inward migration of cations as a mirror image of the oxidation-reaction path operating under dry conditions (Cooper et al. 1996). As no migration of divalent cations was observed, this mechanism does not operate here.

From a microscopic point of view, there are several ways to incorporate these protons into the Eburru melt. If molecular hydrogen (H2) solubility in the melt is sufficiently high, proton incorporation can be achieved by H2 dissolution at the vapor- melt boundary together with H2 diffusion within the melt plus H2 breakdown by reduction of Fe³⁺ at the reduction front, re- sulting in the formation of two Fe²⁺ and two protons (see Fig.

5). In a peralkaline melt such as Eburru, Fe³⁺ (in tetrahedral coordination) is charge-compensated structurally by alkalis (Dickenson and Hess 1986). Therefore, to account for the in-

0.02 0.03 0.04 0.05 0.06 0.07 0.08

0 100 200 300 400 500 600 700 800 900

Molar fraction of oxides

0 0.005 0.01 0.015 0.02 0.025 0.03 Na2O FeO Al2O3

K2O H2O

Redox front

Molar fraction of H2O

FIGURE 3. EMPA and FTIR profiles from sample no. Eb/5 (3 h). The zero value on the X- axis corresponds to the gas-melt interface. From this interface toward the sample interior, a strong Na decrease is observed. No clear variation of other elements can be identified except for water, shown as H₂O on the right axis. Note the difference in position of the redox front and the Na diffusion front. In 3 hours, molecular H₂O should have migrated over 220 mm (after Zhang et al. 1991), which correspond to the fronts of both Na migration and dehydration. The observed variations in Fe content are within the random variability observed across the whole profile.

GAILLARD ET AL.: REDOX EXCHANGES BETWEEN H₂ AND IRON IN MELTS 313

crease of OH content accompanying reduction of Fe³⁺ by H2, we write [in the notation of Hess (1980)]:

H2 + 2 NaFe³⁺O2 + 2 SiOSi Æ 2 NaOH + 4 (Fe²⁺)0.5OSi (1) accompanied by a structural partioning of water:

NaOH + SiOSi ´ NaOSi + HOSi. (2)

If the rate of reduction is controlled by solubilization of molecular H2 at the vapor-melt interface, the growth of the re- acted layer should depend linearly on time. However, as the process shows a parabolic dependence on time, diffusion within the melt must be rate limiting. Diffusion of molecular H2 in AOQ is 3 to 4 orders of magnitude faster than the growth rate of the reduction rim in Eburru. Either: (1) H2 diffusion in the AOQ and Eburru melts are very different; (2) H2 diffusion is rate-limiting, but additional processes delay the reduction rate of the Eburru melt; or (3) H2 is not directly involved in the reaction and diffusion of H2 is not rate limiting.

In the first case, no specific investigation on the composi- tional dependence of D_H2 is available. Hence, we cannot argue in favor of variable diffusion rates for hydrogen. However, many studies devoted to diffusion properties of neutral molecules in silicate melts [He, Ne, Ar, H2O… see Watson (1994)] showed that the smaller the molecule is, the weaker is the composi- tional dependence of the diffusion coefficient. For He whose size is close to that of H2, DHe varies by much less than one log unit between basalt and rhyolite (Watson 1994) whereas we observed a difference of 3 log unit between D_H2 in AOQ and the reaction rate in Eburru. We therefore consider that even if D_H2 can be different in the AOQ and Eburru melts, this differ- ence cannot explain the contrast in velocity between D_H2 in AOQ and the reaction rate in Eburru.

For the second case (Fig. 5, mechanism 2), Crank (1975) gave an approximate solution for the growth of the reaction layer controlled by the dissolution rate of a reactive gas.

X = [(2D_H2·S_H2·t)/(2 CFe3+)]^1/2 (3) where D, S, and C are, respectively, the diffusion coefficient, solubility, and concentration of the subscripted species, and X gives the thickness of the reacted layer as a function of time, t.

This equation may provide an explanation for the difference between the reaction rate and the diffusivity of the mobile spe- cies. In the present case, if the ratio S_H2/2 CFe3+ <<1 but still large enough that transport of molecular H2 dominates the re- duction process, the growth of the reduced rim can be much slower than diffusion of H2 while still obeying a diffusion-lim-

0 5.10^-08 1.10^-07 1.5.10^-07 2.10^-07 2.5.10^-07

0 2000 4000 6000 8000 10000 12000

Run duration (sec)

Square of thereduction front advancement

FIGURE 4. Identification of the kinetic law of Fe³⁺ reduction by hydrogen in the Eburru melt. The linear relationship between the square of the position and run duration indicates a diffusion-limited process.

H₂+ 2 Fe

=>

2 H⁺+ 2 Fe²⁺

3+

Fe -

=>

Fe²⁺

+ e 3+

Fe³⁺+ e-

=>

Fe2+

e^- H⁺ O^2- ne- Mⁿ⁺

H₂+ O^2- H H₂is insoluble in the melt

2O + 2e^-

1

H₂is soluble and reactive in the melt

H₂ 2

H₂ 2 H H₂is insoluble in the melt +

mobile proton ⁺+ 2e^-

3

FIGURE 5. Flow chart illustrating the different likely mechanisms for Fe³⁺ reduction in the Eburru melt. All of these mechanisms must satisfy a required increase of the cation/anion ratio: (1) Molecular H2 is not soluble in the melt, therefore reduction proceeds through O extraction, which is rate-limited by inward migrations of cations or electronic species; this scenario is the mirror image of the oxidation mechanisms observed by Cooper et al. (1996). (2) Adsorption, diffusion, and reaction of molecular H2 with Fe³⁺ in the melt. (3) Molecular H2 is not soluble in the melt but inward migration of protons and electronic species may occur producing the reduction.

ited law. The lack of an H2 signature in the infrared spectra of the Eburru glass is consistent with this solubility criterion.

In the third case, if we consider that molecular H2 is so in- soluble in the melt that there are insufficient H2 species avail- able to reduce the Fe³⁺, other water-derived species must be rate limiting. Thus, molecular H2 may react with Fe³⁺ to pro- duce Fe²⁺ and OH groups at the gas-melt interface, as shown in Equation 1. Then the redox potential is conveyed toward the interior of the sample by coupled transport of protons and elec- tronic species (see Fig. 5, mechanism 3). In Equation 1, H2

should then be replaced by H⁺ and electrons e^– or electron holes h⁺. Because Fe-bearing glasses and melts are polaron-type semi- conductors (Cooper et al. 1996, and references therein), such migration of cations in a redox gradient is charge-compensated by electronic migration. Diopside (Hercule and Ingrin 1999) and olivine crystals (Kohlstedt and Mackwell 1998) have been shown to facilitate incorporation of protons coupled to out- ward migration of electron holes caused by reduction of Fe³⁺. In that case, the growth rate of the reduced rim is controlled either by proton or electronic mobilities. According to Schmalzried (1981), the width of the reacting layer is directly proportional to the square root of time, t, to the diffusion coef- ficient of the rate-limiting species i (H⁺ or electronic species), and to the Gibbs free energy of the ongoing reaction normal- ized by the temperature, T:

X = [ 2 DI·DG / (RT)·t ]^1/2 (4) In general, proton diffusion in SiO2-rich silicate melts is not regarded as a likely mechanism (Zhang et al. 1991; Nowak and Behrens 1997). Although Behrens and Nowak (1997) ruled out the possibility of proton migration in Fe-free SiO2-rich melt, incorporation of Fe may introduce electron holes (Fe³⁺) that provide charge compensation for proton jumps. Also, accord- ing to Stanton et al. (1990) (see also the review of Dingwell 1995), a mobile, positively charged H-bearing species can domi- nate the transport of water-derived species in low-H2O melts (<0.8 wt% H2O; Stanton et al. 1990). This mobile, water-de- rived species could be NaH2O⁺, H3O⁺, or H⁺. We can exclude NaH2O⁺ in our experiments because the chemical profile for Na does not extend all the way to the redox front. A mass- balance calculation, based on the Mössbauer data (Fe³⁺/Fetot) and FTIR spectra (OH) close to the reduction front, indicates variation of Fe³⁺ and OH concentration that reasonably matches with proton rather than H3O⁺ incorporation (one OH band cre- ated for the reduction of one Fe³⁺). Therefore, the parabolic rate constant extracted from Figure 4 should reflect either pro- ton or electron migration rates, depending on which is the slowest.

Na migration

The migration of Na seems to be decoupled kinetically from the redox interaction between H2 and iron. A rapid decrease in the Na content (of ~7%) is observed within the first 30 min of exposure to an H2 gas. We interpret these results as the conse- quence of the redox partitioning of Na between melt and gas already observed under H-free conditions (see Georges et al.

2000). The f_O2 during the beginning of the runs is very reduc- ing so that it causes a displacement toward the RHS of the

following equilibria:

Na2O^melt ´ Na2gas + ¹/2O2gas (5) As we did not observe any Na gradient in sample no. Eb/1, the above equilibria should be achieved very rapidly consis- tent with very high Na diffusion rate in peralkaline melts (Henderson et al. 1985). Molecular O2gas is not stable, as H2O will be produced instead at the melt/gas interface by oxidation of H2 by Na2O. This oxidation may effect a small, transient change in the f_O2 of the gas near the surface of the melt. In sample no. Eb/4 and no. Eb/5 (Fig. 3), as well as dehydration of the melt, we observed chemical migration of Na. The re- lease of water into the gas phase should increase the f_H2O/f_H2 ratio and therefore increase the f_O2 in the gas. Equation 5 is therefore displaced to its LHS and Na migration from the gas toward the melt occurs. Figure 3 shows that the position of both Na migration and dehydration fronts nearly corresponds to the calculated position of a front of molecular H2O migra- tion [calculated after Zhang et al. (1991), see caption of Fig.

3]. Therefore, we anticipate that migration of Na from the gas to the melt operated at a rate controlled by melt dehydration (i.e., H2O mobility) and not by Na diffusion in the melt.

Comparison with previous work

The kinetics of Fe oxidation-reduction in anhydrous basal- tic melts has been demonstrated to be rate-limited by diffusion of divalent cations (Cooper et al. 1996; Cook and Cooper 2000).

In contrast, we provide clear evidence that, for low H2O melts, molecular hydrogen or coupled proton-electronic species are the mobile species controlling the advance of the reduction front. The kinetics of change in Fe³⁺/Fe²⁺ is thus much faster in H-bearing systems than in anhydrous systems (representing the several orders of magnitude difference between Ca²⁺ and H2, H⁺ diffusion rates). Given that H-bearing species are always present in natural magmas, modification of the Fe redox state is most likely to occur following the mechanisms we have iden- tified in this study. It should be noted, however, that for high- H2O, low-Fe melts, Gaillard et al. (2002) have not detected any coupled proton/electronic species motion. The reason for this discrepancy may be due to reduced electronic conductiv- ity in Fe-poor melts and/or high concentration of water in the melts that may affect both diffusion and solubility of H2.

CONCLUDINGREMARKSANDPERSPECTIVES The major points from this study are: (1) reduction of Fe³⁺

in the melt by H2 operates by the progression of a redox front that is accompanied by an increase in the concentration of OH- groups; (2) H2 diffusion in silicate melts is much more rapid than the observed rate of reduction of Fe³⁺; (3) the velocity of the redox front within the melt is faster than the mobility of molecular H2O, suggesting that various hydrogen-bearing spe- cies with different diffusivities may exist in a silicate melt; and (4) exchange of Na between the gas and the melt occur at a very high rate and are extremely sensitive to the redox conditions.

Understanding the response of a magma to changes in its environment, such as during degassing or exchanges during mixing or interaction with surrounding solids requires the

GAILLARD ET AL.: REDOX EXCHANGES BETWEEN H₂ AND IRON IN MELTS 315

knowledge of the identity and the mobility of all H-derived species. In this study, we have shown that redox exchanges between H and Fe causes incorporation of OH groups at a rate different from H2O migration. We have also shown that the H content and Fe³⁺/Fe²⁺ ratios in melt are strongly correlated. We should thus question the robustness of Fe³⁺/Fe²⁺ of melts as an indicator of pre-eruptive oxygen fugacity because these melts may have undergone degassing (Christie et al. 1986). Further studies, however, are clearly needed to quantify the relation- ship between Fe redox state and H incorporation/extraction for a range of melts compositions and experimental conditions that permit direct comparison with natural magma processes.

ACKNOWLEDGMENTS

Interpretations of the experimental data were greatly enhanced by discus- sions with Reid Cooper and Rebecca Everman. Julian Mecklenburgh made a useful informal review of the final version. Ray MacDonald, who kindly sup- plied the Eburru glass, is also thanked. Hans Keppler and Max Wilke are ac- knowledged for their reviews. Mike Toplis is greatly thanked for the editorial work.

REFERENCESCITED

Behrens, H. and Nowak, M. (1997) An experimental investigation on diffusion of water in haplogranitic melts. Contribution to Mineralogy and Petrology, 126, 365–376.

Chekhmir, A.S, Persikov, E.S., Epel´Baum, M.B., and Bukhtiyarov, P.G. (1985) Experimental investigation of the hydrogen transport through the model mag- matic melt. Geokhimya, 5, 594–598.

Christie, D.M., Carmichael, I.S.E., and Langmuir, C.H. (1986) Oxidation state of mid-ocean ridge basalt glasses. Earth Planetary Science Letters, 79, 397–411.

Cook, G.B. and Cooper, R.F. (2000) Iron concentration and the physical processes of dynamic oxidation in alkaline earth aluminosilicate glass. American Miner- alogist, 85, 397–406.

Cooper, R.F, Fanselow, J.B, and Poker, D.B (1996) The mechanism of oxidation of a basaltic glass: Chemical diffusion of network-modifying cations. Geochimica et Cosmochimica Acta, 60, 3253–3265.

Crank, J. (1975) Mathematics of diffusion. Oxford University Press, U.K.

Dickenson, M.P. and Hess, P.C. (1986) The structural role and homogeneous redox equilibria of iron in peraluminous, metaluminous and peralkaline silicate melts.

Contributions to Mineralogy and Petrology, 92, 207–217.

Dingwell, D.B. (1995) Relaxation theory in silicate melts. In J. Stebbins, P.F.

McMillan and D.B. Dingwell, Eds., Structure, Dynamics and Properties of Sili- cate Melts, 32, 23–48. Reviews in Mineralogy, Mineralogical Society of America, Washington, D.C.

Gaillard, F., Scaillet, B., Pichavant, M., and Beny, J.M. (2001) The effect of water

and fO2 on the ferric-ferrous ratio of silicic melts. Chemical Geology, 174, 255–

273.

Gaillard, F., Scaillet, B., and Pichavant, M. (2002) Kinetics of iron oxidation-reduc- tion in hydrous silicic melts. American Mineralogist, 87, 829–837.

Georges P., Libourel G., and Deloule E. (2000) Experimental constraints on alkali condensation in chondrule formation. Meteorites and Planetary Sciences, 35, 1183–1186.

Henderson, P., Nolan, J., Cunningham, G.C., and Lowry, R.K. (1985) Structural controls and mechanisms of diffusion in natural silicate melts. Contribution to Mineralogy and Petrology, 89, 263–272.

Hercule, S. and Ingrin, J. (1999) Hydrogen in diopside: Diffusion, kinetics of ex- traction-incorporation, and solubility. American Mineralogist, 84, 1577–1587.

Hess, P.C. (1980) Polarization model for silicate melts. In R.B. Hargraves, Ed., Physics of magmatic processes, 3–48. Princeton University Press, New Jersey.

Ingrin, J. and Skogby, H. (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. European Journal of Mineral- ogy, 12, 543–570.

Johnson, M.C., Anderson, A.T., and Rutherford, M.J. (1994) Pre-eruptive volatile contents of magmas. In M.R. Carroll and John R. Holloway, Eds., Volatiles in magmas, 30, 281–323. Reviews in Mineralogy, Mineralogical Society of America, Washington, D.C.

Kohlstedt, D.L and Mackwell, S.J. (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Zeitschrift für physikalische Chemie, 207, 147–162.

Macdonald, R. and Bailey, D.K. (1973). The chemistry of the peralkaline oversatu- rated obsidians. U.S. Geological Survey Professional Papers 440-N, Part 1, 1–

37.

McCammon, C. (1994) A Mössbauer milliprobe: Practical considerations. Hyper- fine Interactions, 92, 1235–1239.

McCammon, C.A., Chaskar, V., and Richards, G.G. (1991) A technique for spatially resolved Mössbauer spectroscopy applied to quenched metallurgical slags.

Measurement Science Technology, 2, 657–662.

Nowak, M. and Behrens, H. (1997) The mechanisms of water diffusion in polymer- ized silicate melts. Contribution to mineralogy and petrology, 126, 377–385.

Schmalzried, H. (1981) Solid state reaction, Verlag Chemie, Weinheim.

Schmidt, B.C., Holtz, F.M., and Beny, J-M. (1998) Incorporation of H2 in vitreous silica, qualitative and quantitative determination from Raman and infrared spec- troscopy. Journal of Noncrystalline Solids, 240, 91–103.

Stanton, T.R., Tyburczy, J.A., Holloway, J.R., and Petuskey, W.T. (1990) Electro- migration of water in silicate glass: Resolution of the valence charge of the diffusing species. EOS Transactions AGU, 71, 652.

Stolper, E. (1982) The speciation of water in silicate melts. Geochimica et Cosmochimica Acta, 46, 2606–2620.

Watson, E.B. (1994) Diffusion in volatile-bearing magmas. In M.R. Carroll and John R. Holloway, Eds.,Volatiles in magmas, 30, 371–411. Reviews in Miner- alogy, Mineralogical Society of America, Washington, D.C.

Zhang, Y., Stolper, E.M., and Wasserburg, G.J. (1991) Diffusion of water in rhy- olitic glasses. Geochimica Cosmochimica Acta, 55, 441–456.

MANUSCRIPTRECEIVED NOVEMBER 6, 2001 MANUSCRIPTACCEPTED OCTOBER 8, 2002 MANUSCRIPTHANDLEDBY MICHAEL TOPLIS